BSTJ 2: 4. October 1923: Transatlantic Radio Telephony. (Arnold, H.D.; Espenschied, Lloyd)

Transatlantic Radio Telephony 1

By H. D. ARNOLD and LLOYD ESPENSCHIED

Synopsis: The first transmission of the human voice across the Atlantic-
was accomplished by means of radio in 1915. Since that time substantial
progress has been made in the art of radio telephony and in January of
this year another important step was taken in the accomplishment of trans-
oceanic voice communication. At a prearranged time telephonic messages
were received in London from New York clearly and with uniform intensity
over a period of about two hours.

These talking tests were part of a series of experiments on transatlantic-
telephony which are now under way, the results of which to date are re-
ported in this paper.

A new method of transmission, radiating only a single side-band, is being
employed for the first time. As compared with the ordinary method of
transmission, this system possesses the following important advantages:

The effectiveness of transmission is greatly increased because all of the
energy radiated is effective in conveying the message; whereas in the ordinary
method, most of the energy is not thus effective.

The stability of transmission is improved.

The frequency band required for transmission is reduced, thus conserving
wave length space in the ether and also simplifying the transmitting antenna
problem.

An important element of the high-power transmitter is the water-cooled
tubes, by means of which the power of the transmitted currents is amplified
to the order of 100 kilowatts or more. The direct-current power for these
tubes is supplied from a 60-cycle, a-c. source through water-cooled rectifier
tubes.

A highly selective and stable type of receiving circuit is employed.
Methods and apparatus have been developed for measuring the strength
of the electromagnetic field which is delivered to the receiving point and
for measuring the interference produced by static.

The transmission tests so far have been conducted on a wave length of
5260 meters (57,000 cycles per second). The results of the measurements
during the first quarter of the year on the transmission from the United
States to England show large diurnal variations in the strength of the
received signal and in the radio noise strength, as is to be expected, and
correspondingly large diurnal variations in the ratio of the signal to noise
strength and in the resulting reception of spoken words. Also, the measure-
ments, although as yet incomplete, show a large seasonal variation.

The character of the diurnal and seasonal variations is clearly indicated
in the figures. The curves present the most accurate and complete data of
this kind yet obtained.

ON January 15, of this year, a group of about 60 people gathered
in London at a prearranged time and listened to messages
spoken by officials of the American Telephone and Telegraph Com-
pany from their offices at 195 Broadway, New York City. The
transmission was conducted through a period of about two hours,
and during this time the words were received in London with as much
clearness and uniformity as they would be received over an ordinary
wire telephone circuit. During a part of the time a loud speaker

'This caper, with the exception of the Appendix, was presented at the Annual
Convention of the A. I. E. E., Swampscott, Mass., June 26-2), 1923, and was printed
in the Journal for August, 1923.

116

TRANSATLANTIC RADIO TELEPHONY 117

was used in connection with the receiving set, instead of head receivers.
The reporters present easily made a transcription of all the remarks,
both with head sets and with the loud speaker.

These tests were made possible by cooperation between the engineers
of the American Telephone and Telegraph Company and the Western
Electric Company, and the engineers of the Radio Corporation of
America and its associated companies. The sending apparatus was
installed in the station of the Radio Corporation of America, at
Rocky Point, L. I., in order to make use of that company's very
efficient multiple-tuned antenna. The receiving apparatus was
installed in the buildings of the Western Electric Company, Ltd.,
at New Southgate, England.

This was not the first time speech had been transmitted from
America to Europe. Transatlantic telephony was first accomplished
in 1915, when the American Telephone and Telegraph Company
transmitted from the Navy station at Arlington, Va., to the Eiffel
Tower, Paris. In these earlier tests, however, speech was received
in Paris only at occasional moments when transmission conditions
were exceptionally favorable. The success of the present tests
indicates the large amount of development which has been carried
out since this first date.

The recent talking tests were carried out as part of an investigation
of transatlantic radio telephony. This investigation is directed
at determining (1) the effectiveness of new methods and apparatus
which have been developed for telephonically modulating and trans-
mitting the large amounts of power necessary for transoceanic opera-
tion, (2) the efficacy of improved methods for the reception of this
transmission and for so selecting it as to give an extremely sharp
differentiation between the range of frequencies transmitted and all
the frequencies outside of this range; and (3) determining the trans-
mission characteristics for transatlantic distances and the variation
of the characteristics with the time of day and the season of the year,
including the measurement of the amount of static interference.

The tests are being continued, particularly as regards the study
of transmission efficiency.

Single Side-Band Eliminated Carrier Method

of Transmission

The method of transmission used in these experiments is what we

know as the single side-band eliminated carrier method 2 . With this

2 For a more complete exposition of this method see U. S. patent No. 1449382
issued to John R. Carson to whom belongs the credit for having first suggested it.
Also see Carson patents Nos. 1,343,306 and 1,343,307.

118 - BELL SYSTEM TECHNICAL JOURNAL

method, the narrowest possible band of wave lengths in the ether is
used, and all of the energy radiated has maximum effectiveness in
transmitting the message.

As had been pointed out in other papers 3 , when a carrier is modulated
by telephone waves, the power given out is distributed over a fre-
quency range, and may be conveniently considered in three parts:
(1) energy at the carrier frequency itself, (2) energy distributed in a
frequency band extending from the carrier upward, and having a
width equal to the frequencies appearing in the telephone waves,
and (3) energy in a band extending from the carrier downward, and
having a similar width. The power at the carrier frequency itself
makes up somewhat more than two-thirds of the total power, even
when modulation is as complete as possible. Furthermore, this
energy can, in itself, convey no message, as is self evident. In the
present method, therefore, the carrier-frequency component is elim-
inated, by methods explained in detail below with the result that a
large saving in power is effected. Each of the remaining frequency
ranges, generally known as the upper and the lower side-band re-
spectively, transmits power representing the complete message. It is
therefore unnecessary to transmit both of these side-bands, so that
in the present method one of them is eliminated. In this way the
transmission of the message uses only half the frequency band re-
quired in the usual method of operation. Similarly the frequency-
band accepted by the receiving set is narrowed to conform to a single
side-band as compared with the usual double side-band reception,
and as a result the ratio of signal to interference is improved. Certain
other advantages of this method will be brought out in the further
discussion.

While these advantages of the single side-band eliminated carrier
method hold good for radio telephone transmission generally, they
become of the utmost importance in transoceanic work, because of
the necessity of conserving power in a system where the transmitting
powers are large, and also because the very limited frequency range
available for long distance transmission makes it imperative that
each part of the range shall be utilized with the greatest of care.
Before discussing the method further, the circuits and apparatus
which are actually used in the tests will be described.

3 "Carrier Current Telephony and Telegraphy" by Colpitts and Blackwell.
Journal A. I. E. E., April, 1921.

"Application to Radio of Wire Transmission Engineering" by Lloyd Espenschied.
Proc. Inst. Radio Engrs., Oct. 1922.

"Relations of Carrier and Side-bands in Radio Transmission" by R. V. L. Hartley.
Proc. Inst. Radio Engrs., Feb. 1923.

TRANSATLANTIC RADIO TELEPHONY . 119

The Transmitting System

The transmitting system is shown in simplified circuit form in
Fig. 1. It is illustrated as grouped into three parts: The low-power
modulating and amplifying stages, shown below in light lines; the
high-power amplifiers, shown in heavy lines above and to the right;
and the rectifier which supplies the power amplifier with high-tension
direct current, shown in the upper left-hand portion of the
diagram.

Referring first to the low-power portion of the system, it will be
seen that the voice currents (from either a telephone line or a local
microphone) are fed into a balanced type of modulator circuit and
are modulated with a carrier current of a frequency of about 33,000
cycles. The operation of the balanced type of modulator in sup-
pressing the unmodulated carrier component is explained in the
Colpitts and Blackwell carrier current paper referred to above. The
result of this modulating action is to produce in the output circuit of
modulator No. 1, modulated current representing the two side-bands,
for example, the upper one extending from 33,300 to 36,000 cycles
and the lower one from 32,700 down to 30,000 cycles. These com-
ponents are impressed upon a band filter circuit which selects the
lower side-band to the exclusion of the upper one and of any remain-
ing part of the carrier, with the result that only one side-band is
impressed upon the input of the second modulator. This second
modulator is provided with an oscillator which supplies a carrier
current of 88,500 cycles. The result of modulation between
the single side-band and this carrier current is to produce a
pair of side-bands which are widely separated in frequency, the upper
one, representing the sum of the two frequencies, extending from
118,500 to 121,200 cycles and the lower one, representing the difference
between the two frequencies, extending from 58,500 down to 55,800
cycles. In this second stage of modulation there is a relatively wide
separation between the two-side bands which facilitates the selection
at these higher frequencies of one side-band to the exclusion of the
other. Another important advantage is that it allows a range of
adjustment of the transmitted frequency without changing filters.
This is accomplished by varying the frequency of the oscillator in
the second step. In the present case, the frequency desired for trans-
mission is that corresponding to the lower side-band of the second
modulator. The lower side-band of from 58,500 to 55,800 is therefore
selected by means of the filter indicated. This filter excludes not only
the other side-band but also any small residual of 90,000-cycle un-

120 BELL SYSTEM TECHNICAL JOURNAL

ifc — ijmsu ►

-►

<\~

ut

9
tn

i±i

_i
o

z

CO

-^005^

fc

1)000^

r-WWV

HT»

/WW- c

^>

T

Hh+

C

3

^GQCmJ^

Tt>

Fig. 1

TRANSATLANTIC RADIO TELEPHONY

121

modulated carrier current which may get through the second modu-
lator circuit if it is imperfectly balanced.

Having prepared at low power the side-band currents of desired
frequency it is necessary to amplify them to the required magnitude
for application to the transmitting antenna. This amplification is
carried out in three stages. The first stage increases the power to
about 750 watts, and is shown in Fig. 1 together with the modulating
circuits. This amplifier employs in its last stage three glass vacuum
tubes rated at 250 watts each and operating at 1500 volts.

The output of the 750 watt amplifier is applied to the input of
the larger-power amplifying system beginning with the 15-kw. ampli-

Fig. 2

fier of Fig. 1. This consists of two water-cooled tubes in parallel,
operating at approximately 10,000 volts. The output of this amplifier
is applied by means of a transformer to the input of the 150-kw.
amplifier which consists of two units of ten water-cooled tubes each,
all operating in parallel at about 10,000 volts.

The high-voltage, d-c. supply is furnished by a large vacuum tube
rectifier unit rated at 200 kw. It employs water-cooled tubes similar

122

BELL SYSTEM TECHNICAL JOURNAL

to those used in the power amplifiers except that they are of the
two-electrode type. The rectifier operates from a 60-cycle, three-
phase supply circuit and utilizes both halves of each wave. The
two sets of rectified waves are combined by means of an inter-phase
reactor which serves to smooth out the resultant current and by
distributing the load between tubes of adjacent phases increases the
effective load capacity of the rectifier. The ripple is further reduced
by the filtering retardation coil and condensers shown.

Fig." 3

Reproductions of the apparatus comprising the transmitter system
as described above are given in Figs. 2, 3, 4 and 5.

Fig. 2 shows the apparatus comprising the low-power stage of the
transmitting system. The right-hand rack contains the two weak-
power modulating units and the two single-side-band selecting filters.
The left-hand rack is the 750-watt amplifier unit. The three radiation-
cooled tubes of 250-watt capacity each will be seen near the top.

TRANSATLANTIC RADIO TELEPHONY

123

Below are the smaller amplifying stages. The power supply board is
shown in the center of the photograph.

Fig. 3 is a side view of the 15-kw. amplifier unit. The face of the
panel from which the control handles protrude is on the left. Mounted
in the cage behind the front panel are two water-jackets for accom-
modating the water-cooled tubes, also a coiled hose for increasing the
electrical resistance of the water supply circuit (the water-cooled
anodes of the tubes being operated above ground potential).

The final amplifier of 150-kw. capacity is shown in Fig. 4. It
comprises two units each of 75 kw. Each unit contains 10 water
cooled tubes which can be seen mounted in their water jackets.
To the right of these units is located the 200-kw. rectifier unit shown
in Fig. 5. The unit contains actually 12 tubes, there being two tubes

Fig. 4

for each of the six half waves. The pancake coils on the top of the
rack are protecting choke coils to guard the transformer secondary
winding against steep wave fronts in case of tube failure.

From the above description it will be understood that the trans-
mitting system is one in which the useful side-band is first developed

124

BELL SYSTEM TECHNICAL JOURNAL

by modulation and filtration at low power and then its power is
built up to a large value in a succession of powerful amplifiers. It

Fig. 5

will be appreciated, therefore, that the large-power amplifiers and in
particular the water-cooled tubes which are their essential elements
represent one of the major problems of the development.

High-Power Tubes

The development of the high-power tubes is described quite fully
in another paper 4 . The present discussion is, therefore, limited to a
few of the outstanding features.

*Bell System Technical Journal, July 1922.

TRANSATLANTIC RADIO TELEPHONY 125

In the design of high-power tubes for use in this system the main
problem is to insure the ready disposal of the large amounts of heat
generated at the anodes. For the conditions of use in the present
type of system where the tube is employed as an amplifier, the power
which must be disposed of as heat at the anode is of the same order of
magnitude as the power which the tube will deliver to the antenna.
In the case of the present equipment, therefore, the tube must be so
designed as to operate continuously with a heat dissipation at the
anode of more than 10 kw. It is obviously difficult to secure so large
a dissipation in a tube enclosed with glass walls, and a tube was
therefore designed in which the anode forms a part of the wall of the
containing vessel and the heat generated in it is removed by circu-
lating water. The tube used is shown in Fig. 6. The lower cylindrical-
portion is the anode which is drawn from a sheet of copper. The

Fig. 6

upper portion is of glass and serves both to support and insulate the
grid and filament elements.

The three principal difficulties met in the construction of these
tubes are the making of a vacuum-tight seal between the copper and
the glass, the provision of adequate means for conducting through the
glass wall the large currents necessary to heat the filament, and the
obtaining of the necessary vacuum for high-power operation.

126 BELL SYSTEM TECHNICAL JOURNAL

The first of these problems was solved by the development of a
new metal to glass seal. In making this seal the glass and metal parts
are brought into contact while hot, the temperature being high enough
for the glass to wet the metal. The part of the metal in contact with
the glass is made so thin that the stresses which are set up when the
seal cools are not great enough to fracture the glass or to break it
away from the metal at the surface of contact. Seals made in this
way are sufficiently rugged to stand repeated heating and cooling
from the temperature of liquid air to that of molten glass without
deterioration.

A seal employing the same principle but different in form is also
used at the point where the leads carrying the filament current pass
through the glass walls of the tube. The lead is made of copper 0.064
in. in diameter and passes through the center of a copper disk, 0.010
in. thick, the joint between the lead and the disk being made vacuum-
tight by the use of a high melting solder. The disk is sealed to the end
of a glass tube which is in turn sealed into the glass wall of the vacuum
tube.

In exhausting the tubes it has been found necessary to subject all
the metal parts to a preliminary heat treatment in a vacuum furnace
during which the great bulk of the occluded gasses is removed. By
this method the time of exhaust can be considerably reduced but the
vacuum conditions to be met are so stringent that the final processes
of evacuation must be carefully controlled and often occupy as much
as twelve hours.

The tubes are operated at a plate voltage of 10,000 volts and are
capable of delivering 10 kw. at this voltage in a suitable oscillatory
circuit. For this performance an average electron current of 1.35
amperes is required. The total electron current that the filament
must be capable of supplying to insure steady operation is about 6

amperes.

When the tubes are used to amplify modulated currents with large
peak values such as are characteristic of telephone signals it is essential
that the maximum electron current through the tube shall be several
times the normal operating current and therefore to insure the neces-
sary high quality of transmission these tubes are operated for tele-
phone purposes with an average output of about 5 kw.

The Receiving System

In the method of transmission ordinarily employed in radio telephony
by which the carrier and both side-bands are sent out from the trans-
mitting station and received at the distant end, detection is readily

TRANSATLANTIC RADIO TELEPHONY

127

accomplished merely by permitting all of these components to pass
through the detector tube. The detecting action whereby the voice-
frequency currents are derived, is accomplished by a remodulation
of the carrier with each side-band.

With the present eliminated carrier method of transmission the
side-band is unaccompanied by any carrier with which to remodulate
in the receiving detector. It is necessary, therefore, to supply the

Loop

f|

» fed V

■-■■irry—

SCHEMATIC OF SINGLE SIDE BAND RECEIVER

- if tapnr

jSfflSr

Fig. 7

detector with current of the carrier frequency obtained from a local
source. Thus, in the present experiments, if a current of the original
carrier frequency, 55,500 cycles, is supplied to the detector it will
remodulate or "beat" with the received side-band of, say 55,800 to
58,500 cycles and a difference-frequency band of 300 to 3000 cycles,
i.e., the voice frequency band will result.

The arrangement actually used, however, is not quite so simple as
this. It is shown schematically in Fig. 7. Reception is carried out
in two steps, the received side-band being stepped down to a lower
frequency before it is detected. The stepping down action is accom-
plished by combining in the first detector the incoming band of 55,800
to 58,500 cycles with a locally generated current of about 90,000
cycles. In the output circuit of the detector the difference-frequency
band of 34,200 to 31,500 cycles is selected by a band filter and passed
through amplifiers and thence to the second detector. This detector
is supplied with a carrier of 34,500 cycles which, upon "beating"
with the selected band, gives in the output of the detector the original
voice-frequency band.

The object of thus stepping down the received frequency is to secure
the combination of a high degree of selectivity with flexibility in
tuning. The high selectivity is obtained by the use of a band filter.

12S

BELL SYSTEM TECHNICAL JOURNAL

It is further improved by applying the filter after the frequency is
stepped down rather than before. To illustrate this improvement
assume that there is present an interfering signal at 60,000 cycles, 1,500
cycles off from the edge of the received telephone band. This is a
frequency difference of about 2]/ 2 per cent; but after each of these
frequencies is subtracted from 90,000 cycles, the difference of 1500
cycles becomes almost 5 per cent. This enables the filter to effect a
sharper discrimination against the interfering signal. Furthermore,
the filter is not required to be of variable frequency as would be the

Fig. 8

case were it employed directly at the received frequency since by
adjusting the frequency of the beating down oscillator the filter is in
effect readily applied anywhere in a wide range of received frequencies.
The receiving method, therefore, enables the filter circuit, and indeed
also the intermediate frequency amplifiers, to be designed for maximum
efficiency at fixed frequency values without sacrificing the frequency
flexibility of the receiving set.

A photograph of the receiving set used in the transatlantic measure-
ments is reproduced in Fig. 8. The signals are received on a square

TRANSATLANTIC RADIO TELEPHONY 129

loop six feet on a side and wound with 46 turns of stranded wire. The
first box contains the beating oscillator and high-frequency detector,
the second box of the filter and amplifying apparatus for the inter-
mediate frequency and the third box the final detector and amplifier.
The shielded box at the left of the picture, which is connected to the
loop by means of leads in the copper tube, is the apparatus for intro-
ducing the comparison signal of known strength into the loop for
measuring purposes. This receiving and measuring set is described
more in detail in a paper by Bown, Englund, and Friis in the "Pro-
ceedings of the Institute of Radio Engineers for April, 1923."

Although it was this very selective and reliable method of inter-
mediate-frequency reception which was used in London, it is quite
possible to receive the single-side-band transmission by means of a
regular heterodyne receiving set. Even a self-regenerative set will
suffice under some conditions. It is necessary, however, to adjust
the frequency of the oscillator very carefully to that of the trans-
mitting carrier frequency, otherwise serious distortion of the received
speech will result. Also it is, of course, necessary that the tuning
be not too sharp if ordinary tuned circuits and not filter circuits are
employed. One might expect that some difficulty would be experi-
enced in maintaining the frequency at the receiving end in sufficiently
close agreement with the sending frequency. In the tests no par-
ticular difficulty was experienced, the oscillators at the two ends
being so stable that only an occasional slight readjustment of the
receiving oscillator was required. With the development of more
stable oscillators, doubtless, the frequency with which readjustments
are required, will be further reduced. If serious distortion of the
received speech is to be avoided the two frequencies must be within
about 50 cycles, an accuracy of 0.1 per cent at 50,000 cycles.

Transmission Advantages of the System

Since the present experiment represents the first use of the single-
side-band eliminated carrier type of system some further discussion
of the characteristics and advantages of the system is appropriate.

The importance of the system in conserving frequency range will
be appreciated when it is realized that the total frequency range
available for transatlantic telephony is distinctly limited. Just
what the most suitable range is has not been accurately determined
but it seems limited to below 60,000 cycles (5000 meters) because of
the large attenuation suffered during the daylight hours by frequencies
higher than this. On the lower end of the frequency scale, transr

130 BELL SYSTEM TECHNICAL JOURNAL

atlantic telegraphy at present pretty well preempts frequencies below
30,000 cycles (10,000 meters). Therefore, for the present at least
transatlantic telephony is limited to a range of some 30,000 cycles.
Now transmission of speech requires as a minimum for good quality
a single-side-band 3000 cycles wide. Allowing for variations and
clearances between channels it is doubtful if the channels could be
made to average closer than one every 4000 cycles for single-side-
band transmission and one every 7000 cycles for the ordinary double-
side-band transmission. This means that even were the whole range
from 30,000 to 60,000 cycles devoted to telephony to the exclusion of
telegraphy, only about four channels could be obtained by the older
methods and some seven by the present one.

It is a rather interesting commentary to note that a somewhat
similar situation as to limitation in frequency range exists in the
case of carrier-current transmission over wires. The transmission
efficiency falls off with increase in frequency and limits the range
of frequencies which can be economically used, in much the same
way as it is limited in long distance radio transmission. It is because
of this limitation in the case of wires and the value which attaches
to conserving the frequency range consumed per message that single-
side-band transmission was first developed for wire carrier current
systems. Its development in wire transmission has been of con-
siderable aid in adapting the method to the present purpose of trans-
atlantic operation.

The second of the outstanding characteristics of the present system
resides in the large power economy which it permits. Transatlantic
telephony requires hundreds of kilowatts of high-frequency power.
Since it is difficult and expensive to produce this power it is important
that every effort be made to increase its efficiency or effectiveness in
transmitting the voice. To illustrate how the present system effects
economies in power, consider the case of a carrier wave completely
modulated by a single frequency tone. In such a completely modu-
lated wave, only 1/3 of the total power contains the message, the
remaining 2/3 conveying only the carrier frequency which can as
well be supplied from an oscillator of small power at the receiving
station. It is obvious, therefore, that by eliminating the carrier
only 1/3 as much power need be used as would be required were all
the elements of the completely modulated wave transmitted. To
realize the maximum advantage of this mode of operation, the system
eliminates the carrier at low power and, thereby, the high-power
apparatus is devoted exclusively to the amplification of the essential
part of the signal.

TRANSATLANTIC RADIO TELEPHONY 131

If, after having suppressed the carrier, both side-bands were trans-
mitted, their reception would require perfect synchronism between
the carrier resupplied at the receiving end and that eliminated at the
sending end, a condition which is practically impossible to meet
without transmitting some form of synchronizing channel, which is,
indeed, much the same as transmitting the carrier itself. If the
receiving carrier is not synchronized, the two side-bands will interfere
with each other upon being detected. By eliminating one side-band,
this interference is prevented and reception may be carried on, using
a locally supplied frequency which is only approximately equal to
that of the suppressed carrier. The two may differ by as much as
50 cycles before the quality of the received speech is greatly impaired.
The importance to the carrier suppression method of eliminating
one side-band will, therefore, be appreciated. The present system
eliminates one side-band while still in the low-power stage. While
it would be possible to do this selecting after they have both been
raised to the full transmitting power, this would require the use of a
filter of high-power carrying capacity, which would make the filter
very costly and also render the system inflexible to change of wave
length. The present system overcomes both of these difficulties by
filtering our one side-band at low-power levels and by the use of the
double modulation method.

Another very important reason for the transmission of a frequency
band as narrow as is possible lies in the difficulty of constructing an
antenna to transmit more or less uniformly at these long waves a
band of frequencies which is an appreciable fraction of the main
carrier frequencies. For example, in the ordinary method of trans-
mission an antenna which was intended to transmit a 30,000-cycle
carrier and its two speech side-bands would need to be designed to
transmit all the frequencies from 27,000 cycles to 33,000 cycles, a band
which is equal to 20 per cent of the carrier frequency. This band is
considerably wider than that given by the resonance curve of a highly
efficient long wave antenna. To accommodate both side-bands would
require flattening out the resonance curve either by damping, which
means sacrifice in power efficiency, or by special design of the antenna,
possibly throwing it into a series of interacting networks and causing
it to become a rather elaborate wave filter. The importance, from
the antenna standpoint, of narrowing the frequency band required to
be transmitted is, therefore, evident.

It is extremely important that the received signal be affected as
little as possible by changes in the transmission efficiency of the
medium. The voice frequency currents produced at the receiving

132 BELL SYSTEM TECHNICAL JOURNAL

end, after detection, are proportional to the product of the carrier
wave and the side-band. If the carrier as well as the side-band is trans-
mitted through the medium, then a given variation in the transmission
efficiency of the medium will affect both components and will change
the received speech in proportion to the square of the variation, as
compared to the first power if only the side-band is transmitted and
the carrier is supplied locally. Thus it will be seen that the omission
of the carrier from the sending end and the resupplying of it from
the constant source at the receiving end gives greater stability of
transmission.

Without discussing the system in further detail the advantages of
it may be summarized as follows:

1. It conserves the frequency (wave length) band required for radio
telephony, which is particularly important at long wave lengths.

2. It conserves power, in that all of the power transmitted is useful
signal-producing power. This is particularly important also in
long distance transmission which requires the use of large powers.

3. The fact that only a single-band of frequencies is transmitted
simplifies the antenna problem at long wave lengths, where the
resonance band becomes too narrow to transmit both side-bands.

4. As compared with a system which eliminates the carrier but
transmits both side-bands the simple side-band system has the
important advantage of not requiring an extreme accuracy of
frequency in the carrier which is resupplied at the receiver. Were
both side-bands transmitted very perfect synchronism would be
required for good quality.

5. It improves the transmission stability of the radio circuit since
variations in the ether attenuation affect only one (the side-band)
of the two components effective in carrying out the detecting
action in the receiver.

6. The receiving part of the overall system has two advantages:

a. It need accept only half of the frequency band which would be
required in double side-band transmission, thereby accepting
only half of the "static" interfering energy.

b. By stepping down the frequency of the received currents and
filtering and amplifying at the low-frequency stage a very sharp
cutoff is obtained for frequencies outside of the desired band and
a very stable and easily maintained amplifying system is obtained.

TRANSATLANTIC RADIO TELEPHONY 133

Study of Transatlantic Transmission

We come now to a consideration of the second major part of the
investigation, namely, that having to do with the transmission of
the waves across the Atlantic. It will be evident, from what has been
said earlier, that the transmission question is essentially one of how
best to deliver, through the variable conditions of the ether to the
receiving station, speech-carrying waves sufficiently free from inter-
ference to be readily interpretable in the receiving telephone. The
transmission efficiency of the medium varies with time of day and
year, and is different for different wave lengths. The interference
conditions are also influenced by these same factors.

Now we can study this transmission medium in much the same way
we would a physical telephone circuit, by putting into it, at the
sending end, electromagnetic waves of a known amount of power
and measuring the power delivered at the receiving end. The inter-
ference at the receiving station likewise may be measured and the
ratio of the strength of the signal waves to the interfering waves may
be taken as a measure of freedom from interference; this in turn
being directly related to the readiness with which the messages are
understood. Accordingly, there has been included as an integral part
of the investigation of transatlantic radio telephony, the development
of suitable methods and apparatus for measuring the strength of the
signal waves and of the interfering waves, as they arrive at the receiv-
ing station. The apparatus 5 employed in measuring the field strength
of the received signals has been outlined above under Receiving
System and need not be gone into further. However, a word of
explanation about the method which is employed in making the
measurement may be helpful. It will be recalled that the specially
designed receiving set is provided with a local source of high fre-
quency from which can be originated signals of predetermined strength.
A measurement of the field strength of a signal received from the
distant transmitter is made by listening first to the distant signal and
then to the locally produced signal, shifting back and forth between
these signals and adjusting the strength of the local signal until the
two are substantially of the same strength. Then, knowing the
power delivered by the local source, the power received from the
distant station is likewise known. The relation between the power
in the input of the radio receiving circuit to the field strength required
to deliver that power is known through the geometry of the receiving

6 It is described in detail in the paper entitled, "Radio Transmission Measurements"
by Bown, Englund, and Friis, Proc. Institute of Radio Engrs., April, 1923.

134 BELL SYSTEM TECHNICAL JOURNAL

antenna (in this case a loop) and, therefore, the measured power of
the signal can be translated directly into the field strength of the
received waves.

The measurement tone signal is transmitted from the Rocky Point
sending station by substituting for the microphone telephone trans-
mitter a source of weak alternating current of about 1/100 watt at
a frequency of approximately 1500 cycles. This tone modulates the
radio telephone transmitter in the same way that voice currents
would and is radiated from the antenna as a single-frequency wave
of 5260 meters (57,000 cycles per second). It, therefore, constitutes
a means of sending out a single-frequency continuous wave for
measurement purposes. At the receiving end this continuous wave
is demodulated to the same tone frequency which it originally had.

For measuring the strength of the received noise, i.e., the radio
frequency currents arising from static or other station interference,
the method is quite similar. In this case, however, the noise received
is so different from that which can be set up artificially in any simple
manner that no attempt is made to compare it directly with a local
noise standard. Instead the volume of the interfering noise is expressed
in terms of its effect in interfering with the audibility of a local tone
signal by measuring the local signal which can just be definitely
discerned through it. This is a threshold type of measurement
which is necessarily difficult to carry out with accuracy. In order to
increase the sharpness of definition of the local signal and to make
it correspond more closely to speech reception the signal tone is
subjected to a continuous frequency fluctuation. The comparison
signal has therefore a warbling tone which occupies a frequency band
not unlike that of the voice. This method of measuring the inter-
ference is discussed in more detail also in the measurement paper
referred to above.

Procedure in Making Transmission Measurements. The three
quantities which are included in the transmission measurements,
namely, the signal strength, the noise strength, and the percentage
of words received correctly, are observed one after another in what
might termed a unit test period. Although the duration of this test
period and the order of making the measurements has been changed
somewhat during the course of the experiments, the following pro-
gram is representative of the conditions under which the data pre-
sented below were taken.

A 25-minute test period divided as follows:

5 minutes of tone telegraph identification signals (for receiving
adjustment purposes).

TRANSATLANTIC RADIO TELEPHONY

135

10 minutes of disconnected spoken words.

10 minutes of a succession of five-second tone dashes separated by
five-second intervals, (for measurement of the received field strength,
the intervals between the dashes being used for throwing on the local
receiving source and adjusting its strength to equal that of the receive
signals by alternately listening to one and then the other).

Transatlantic Radio Transmission Measurements

Diurnal Signal ft Noise Variation

Jan.l-Febr.23, 1925.

Shaded area shows range-

of variation. Heavy linea

avernc».

Signal field Strengths corrected

to 300 Amos. Antenna Current.

Fig. 9

Immediately following this test period the London observers
measured the noise level.

This unit test period was repeated every hour over a period which
varied from several hours to as long as two days' duration. Most of
the test periods ran for about 28 hours, starting about eleven o'clock
Sunday morning and continuing until about three o'clock Monday
morning, London time. During this time the telegraph load through

136

BELL SYSTEM TECHNICAL JOURNAL

the Rocky Point station of the Radio Corporation was sufficiently
light to enable one of the two antennas to he devoted to these experi-
ments. The measurements were started January 1, 1923 and are
still in progress.

At the present time (April) the results for the first three months
of the tests are available. These results are not yet sufficiently
complete nor do they cover a sufficient number of variables in terms

Transatlantic Radio Transmission Measurements

Diurnal Signal & Noise Variation

Feb25-Apr.9, 1925.

Shaded area ehowe range

of wriationi Heavy Unas

average.

Signal Field Strengths correct**

to 500 Amps. Antenna Current

Fig. 10

of time, wave length, etc., to enable any very definite conclusions to
be drawn from them. They do illustrate, however, the usefulness of
the methods employed, and even in their incomplete state show some
factors of considerable interest.

The results of the measurements of received signal, strength and
received noise are given in Figs. 9 and 10. The data have been divided
and plotted in these two sets of curves because the transmission con-
ditions across the North Atlantic appeared to suffer a rather rapid
change about February 23rd. Fig. 9 therefore covers the winter
period from January 1 (when the test started) to February 23; and
Fig. 10 covers the next period from February 25 to April 9.

TRANSATLANTIC RADIO TELEPHONY 137

The curves are plotted between time of day as abscissas and field
strength in microvolts per meter as ordinates. The time during which
darkness prevailed at Rocky Point and at London is indicated by the
block-fills on the time scales. The overlap of these block-fills indicates
the time during which darkness extended over the entire transatlantic
path. For Fig. 9 the darkness-belt is as of February 1 and for Fig. 10
as of March 21. The curves show the mean of the results and also
the boundaries of the maximum and minimum values observed.

Received Signal Strength. The outstanding factors to be noted
concerning the signal strength curves are:

1. The diurnal variations are plainly in evidence. During the first
test period covered by Fig. 9, for example, the field strength varied
in the ratio of the order of 15 to 1 between day and night conditions,
running about 100 microvolts per meter during the night and averag-
ing about 6 microvolts per meter during the day. The diurnal varia-
tion is also to be seen in Fig. 10 although the variations between night
and day transmission are less marked.

The measured daylight values lend support to the Austin-Cohen
absorption coefficient. The average of the observed daylight value
for the period of these tests is between 7 and 8 microvolts per meter,
while the calculated value is 9.5. Concerning the high field strengths
obtaining at night, it should be noted that the maximum observed
value of 237 mie'ro volts per meter does not exceed the value of some
340 microvolts which it is estimated should obtain at London were
no absorption present in the intervening medium, i.e., were the waves
attenuated in accordance with the simple inverse-with-distance law.
While no definite conclusions can yet be drawn from these results as
to the cause of the diurnal variations, this indication that the upper
limit of the variation is the no-absorption condition suggests that
the diurnal fluctuations are controlled by the absorption conditions
of the medium rather than by reflection or refraction effects.

2. An indication of the seasonal variation which apparently occurs
in developing from winter to early spring is found in a comparison of
the signal strength curves of Figs. 9 and 10. On the whole the signal
strength received in the second test period is considerably less than
that received for the first period. This drop in the average of the 24
hours is caused by a large decrease in the night-time transmission
efficiency. The daylight transmission does not change much, but
what little change there is lies in the direction of an increase as the
season advances.

3. A decrease in the transmission efficiency is observed between
the time of sundown in London and sundown in New York, that is,

138 BELL SYSTEM TECHNICAL JOURNAL

during the period when the sunset condition intervenes in the trans-
mission path. This dip is particularly noticeable in the signal strength
curve of Fig. 10. It is not noticeable in Fig. 9, except for the fact that
the rise in signal strength corresponding to night conditions in London
is delayed until the major part of the transmission path is in darkness.
Strength of Received Noise. The variation in the strength of received
noise is shown by the noise curves of Figs. 9 and 10.

1. The diurnal variation of that portion of"the noise which is due
to atmospheric or "static" disturbances is somewhat obscured by
the presence of artificial noise, i.e., noise caused by interference from
other stations. The rise in the noise curve at 12 noon is known to
be due to artificial interference. In general, however, the large noise
values shown to prevail throughout the night in London between
about 6 p. m. and 4 a. m. are known to be due to atmospherics. This
diurnal variation shows up quite prominently in both figures.

The maximum noise is reached at 2 a. m. London time. Up to this
time the night belt extends over London and a sector of the earth
considerably to the east and including Europe, Africa and Asia. The
noise begins to drop off shortly thereafter and reaches its minimum
at sunrise in London. This could be accounted for on the assumption
that the major source of the noise lies considerably to the east of
London and that transmission of the stray electric waves to London
is gradually diminished in efficiency as daylight overtakes the path
of transmission.

2. The seasonal variation, as shown by a comparison of the noise
curve of Fig. 9 with that of Fig. 10, is not so great as is the case with
the transmission efficiency of the signal. However, the noise level is
noticeably higher during the second period of the tests, as shown
by the average curve of Fig. 10, particularly during the night when the
maximum noise obtains.

This indicates that the noise is largely of continental origin lying
to the east or south east of London which is in agreement with rough
observations made by means of a loop and suggests that the employ-
ment of directional antennas would be of considerable advantage. It
is expected to include such antennas in the further measurement work.

In connection with these noise curves it should be noted that what
they represent is in reality the strength of a local warbling tone-signal,
expressed in terms of equivalent field strength in microvolts, which
which is just definitely audible through the noise. The actual value
of the noise currents, were they measured by an integrating device such
as a thermocouple, for example, would be a number of times larger
than indicated.

TRANSATLANTIC RADIO TELEPHONY 139

Ratio of Signal to Noise Strength; Words Received. The noise curve
of Fig. 9 and that of Fig. 10 can, therefore, be read as "The strength
of the signal tone which can just be heard through the noise." It can,
therefore, be directly compared with the signal curve itself and the
difference between the two curves is a measure of the level of the actual
signal strength above that which would just permit of the signals being
heard. Actually, the difference between the two curves, as shown in
the figures, is proportional to the ratio of the signal to the noise
strength, because the curves are plotted to a logarithmic scale.

This signal to noise ratio is plotted in Fig. 11 for the test period
which corresponds to Fig. 9, and Fig. 12 for the test period which
corresponds to Fig. 10. These ratio curves are derived by going back
to the original data and taking the ratio for each unit measurement
period and spotting it upon the chart as shown by the black points.
An average is taken of the points for each hour of the 24-hour period
as shown by the circle points. The dash line curves of Figs. 11 and 12,
therefore, trace the average diurnal variation of signal to noise ratio.
These curves show :

1. That the signal to noise ratio reaches its minimum during the
time when the sunset period intervenes between London and New
York.

2. During the night in London the ratio increases more or less con-
tinuously and reaches a maximum around the time of sunrise in
London.

3. During the course of the daylight period in London the ratio
starts out high and drops rather rapidly during the forenoon and
assumes a more or less constant intermediate value during the after-
noon until sundown. It is during this afternoon period in London
that the business hours of the day in London and New York coincide,
so that this is the most important period from a telephone com-
munication standpoint.

The drop in the very low ratios obtaining in London in the early
evening is due to the fact that an increase in noise occurring at this
time is accompanied by a decrease in transmission efficiency from
America. This may readily be seen by referring to Fig. 10. The
noise increases as the night belt, proceeding westward, envelops Eng-
land and improves the transmission of atmospherics, which arise
possibly in continental Europe, Asia and Africa. As the shadow wall,
proceeding westward, intervenes between England and America, the
transmission efficiency of the desired signals from America drops and
it is not until the night belt extends as far west as America that the
transmission efficiency improves sufficiently to overcome the dis-

140

BELL SYSTEM TECHNICAL JOURNAL

advantage in London of the large noise values which night there had
brought on. Conversely, the high signal to noise ratio, obtaining at
about sunrise in London, appears to be due to the fact that as the
termination of the night belt, moving westward, intervenes between
England and the source of atmospherics to the east, the noise level
drops rapidly and has reached low values by the time sunrise arrives
in London. At this time, however, darkness still extends to the west

TRANSATLANTIC RADIO TRANSMISSION MEASUREMENTS

Diurnal Variations of Signal to Noise Ratio

Jan.l-Feb.23. (923.

Each circle i

arerage o

r*su

H of twti for thit hour

soo

-A

s

r

V

WO

/

\

'

<

|

V

'

; c

.

S L

A

s

,

i

1

S

f

^fr---

i

C^

1

,

—

^ —

Jz

1

'

j.

,_

v .Lyf

?

i

1

*

|

M.d

iqhtl

5unri

t

1

2M 1 2

\

|

t

s t a 9 » ii

H

ris

tTS^i

^ra*

) (1 tZM 1 2 i A

1

Fig. 11

and the transmission efficiency from America is at its maximum. It
is, therefore, due to this interplay between these two factors, signal
strength and noise strength, controlled very largely by the transition
periods between day and night, that the signal to static ratio varies
diurnally in the manner pictured in Figs. 11 and 12.

Concerning seasonal variation, shown by a comparison of Figs.
11 and 12, the following can be said: The dimunition in signal-to-
noise ratio in the second test period as compared with the first is
caused by the fact that the signal strength has decreased and at the
same time the noise has somewhat increased. There is just one other

TRANSATLANTIC RADIO TELEPHONY

141

point that concerns the dip in the ratio occurring at night in London
between 12 midnight and 3 a. m. This dip is due to an increase in the
noise which occurs around 2 a. m. (A further reduction during this

Transatlantic Radio Transmission Measurements

Diurnal Variations of Signal to Noise Ratio

Feb. 25- Apr. 9,1923.

Each drtkisaveraoi of results of tests for that hour

CO

—

50

^5fc-1

V-

s:

!

/

1

/

to

—

1-

L

■

4

/

_u

—

<

s

►— 1

1

—
—

PC

— 1

►

H

t —

■ —

-

^—

1
f

— i

— i

-

m

—

—

.

\

/

1

Suisot "

IUI234S678910I

« 9 » It fiH 1 t 5 4 S '

1 BtorVWiTIm* I

Ha

I

mqht

Sunrise
| 3 .5 6 7 8 9 10 11 QM

9(0 II 121 234367

Fig. 12

time, and one which extends the time of minimum ratio from sundown
on through the night until 2 a. m. is shown by the April measurements
which time has not permitted including in the curves).

During each test period lists of disconnected words were spoken
over the systems. As an approximate and easily applied method of
indicating the talking efficiency of the circuit, note was made of the
percentage of the words which were correctly received.

The curves of Figs. 13 and 14 show the manner in which the per-
centage of the words which were correctly received varies through
the 24 hours. Each point corresponds to the percentage of words cor-
rectly received during one unit test period. In many of these tests
the interference was noted to be caused by radio telegraph stations,
and the data in which the interference is of this character, in so far as
identified, are indicated by the triangular dots. It will be seen that
most of the poor receptions were due to this cause. Especially is
this true of tests at 12 noon at which time severe interference from
sources local in London was experienced. The circle points are the

142

BELL SYSTEM TECHNICAL JOURNAL

average of results for each hour's tests. The dash line curve is a
smoothing out curve of these points.

It is interesting to note that these curves of actual word count
conform very well in general_shape with those of Figs. 11 and 12
which also really measure receptiveness although in a less direct
manner. Reception is best during the late night and early morning,

Transatlantic Radio Transmission Measurements

Diurnal Variation of Words Understood

Jan. I -Feb. 23, 1923.

Each circle Is average of all teals for that hour including
triangular points. The latter ere known to be cases in
which low percentage is due to unnatural causes.

»0

80
70
60
50

40
30
20
10

•

_

^

*

— *• — J— 1

***

•\

■

/

^

i

!

\

~~

1

1

^'

s

(

\

s

1 1

•

s

/

N.

11

Mil

ninht

rift

12

M 1 2 3 '

5 6 7 1; 9 10 1

121 23456789 10 11 KM

8 9 10 11 I2M 1 2 3 4 3 6 7 8 9 10 II 12 12 3 4 5 6 7

Fig. 13

drops off during the day, reaching a minimum during the evening.
Furthermore, the night reception is shown to be considerably better
for the January-February period than for the February-March period.
The curve of Fig. 14 corresponds quite closely with that of Fig. 12.
The curve of Fig. 13 does not show as much of a peak as does that of
Fig. 11 which is, of course, due to the fact that above a certain ratio
the percentage of words understood is high and cannot rise above 100
per cent.

r ••

Conclusion

As has been indicated this is a report of work which is still in progress.
To date :

A new type of radio telephone system affording important ad-
vantages for transatlantic telephony has been developed and put
into successful experimental operation across the Atlantic.

Sustained one-way telephonic transmission has been obtained across
the Atlantic for the first time by means of this system.

TRANSATLANTIC RADIO TELEPHONY

143

The advantages of this system which had been anticipated, par-
ticularly, in respect to economies of power and wave lengths, have
been realized. Furthermore, it has been demonstrated that the
high-power water-cooled vacuum tubes which have seen their first
prolonged operation in this installation are admirably adapted for
use in high-power radio installations and particularly for use as high

Transatlantic Radio Transmission Measurements

Diurnal Variation of Words Understood

Feb. 25 -April 9, 1925.

Each circle a average of Bll teete for that hour including TTIrr A A

triangular pointe. The latter are known to be caeea in which -Tig. 1 *
low percentage ia due to unnatural cauaee.

power amplifiers, in the type of system we have described. Also, the
method of reception has proved itself to be eminently satisfactory for
use with the single side-band type of transmission and to possess
important advantages for radio telephony in respect to selectivity and
amplification.

Methods have been developed for measuring the strength of the
received signals and the strength of the received interfering noise and
these methods have been successfully applied in the initiation of a
study of the variations to which transatlantic transmission is subject.

The results of the transmission measurements show that, at 5000
meters, the diurnal variations are large, as was to be expected, and
give evidences of a large seasonal variation which was, indeed, also to be
expected. The results indicate that it will probably be desirable to
use a wave length longer than 5000 meters. The measurements are
now being made to include the longer wave lengths.

APPENDIX Added September 23, 1923
The results of the transmission measurements from January through
August are now available and are summarized in the curves following:

BELL SYSTEM TECHNICAL JOURNAL

&MT.EM 2 4 6 8 10 12 2 4 6 B 10 I2 M
ELST.7 9 11 13570 11 I 357

8 10 12 2 4 6 8 10 I2 M
113579 II 1357

EST.7 9
APWfc AUGUST

Transatlantic Radio Transmission Measurements
Monthly Averages of Diurnal Variations in Signal to Noise Ratio for 1923. Trans-
mission from Rocky Point to London on 57,000 Cycles (5,260 Meters). Measurements
on Loop Reception. Curves Corrected to 300 Amperes Antenna Current